Very interesting pictures. I have a few comments to make.
Photo title: Misalignment at 36 inches segmented bends.
API Std. 5L and subsequent 5LX's allow an out of roundness of 1% of the pipe diameter. 36 inches is roughly 910 millimeters, so 1% is 9,1 millimeters, roughly 3/8 of an inch. Too much of a distance.
Back in my days of erector engineer, when I was in charge of the erection of a 24 inches natural gas pipeline, we used the well known strongbacks and wedges to have the two welding ends rounded up and facing perfectly each other. After this the weld was made, not before. This procedure was time consuming, of course; but more time consuming is to fix up the weld shown on the photo after it has cracked. I'm talking of more than forty years ago, nowadays there are devices that face two pipe welding ends in a few minutes.
Photo title: Improper weld transitions.
Even if approved by B.31.8 I don't like the transition shown in the sketch. At the gas pipeline erection I mentioned the transition of the thicker wall started right at the bevel root and continued in a 30° slope until it reached the pipe inside diameter.
Photo title: Backweld quality.
At the gas pipeline erection someone proposed to repair the weld first pass, when necessary, by backwelding. The client's inspector refused and I, the contractor, agreed (from time to time the client inspector and the contractor agree). Look on the photo at the guy who's backwelding. He's far from being comfortable, for sure. Can someone demand from him a first quality weld? Not me.
At the mentioned job, the few weld defects shown by Rx were removed by grinding and the weld remade.
Photo title: Low strength fitting. Hydrotest failure.
I would say, and insist, that the failure has nothing to see with welds. It looks like a failure due to ductile fragile transition. What was the water temperature? This case deserves to be studied with more detail.
Giovanni S. Crisi
Sao Paulo - Brazil